California Transparency in Frontier AI Act

Statute United States (California) · 2025

SB 53 is the first US law requiring public, standardised safety disclosures from developers of frontier AI models — those trained above 10^26 FLOPs. Large frontier developers must publish a frontier AI framework, publish transparency reports on model release, and report critical safety incidents to California's Office of Emergency Services, with penalties to $1 million per violation. It reaches perhaps a dozen companies and sets the template everyone else will be measured against.

Sectoral regime. It binds the industries below; companies outside them generally fall outside it.

Countries

Where this binds. Each links to the providers apis.io has catalogued there.

Regions

Industries

Each links to that industry as a scored cohort — so the question "how ready is this sector for the regime that governs it?" becomes one you can actually look at.

Implemented by these standards

A regulation is the law; a standard is the machine-readable contract that satisfies it. Almost every regime in this catalog restricts an interface rather than requiring one — where a standard exists, it is the part a provider can actually publish.